Noometry

Model comparison

Gemini 3.8 Flash vs Qwen3-4B

Gemini 3.8 Flash is the stronger model overall, scoring 61.8 to 31.9 on the Noometry Index.

Last verified . 3 shared benchmarks.

Gemini 3.8 Flash Google

61.8

Rank #11 Confirmed

Qwen3-4B Alibaba (Qwen)

31.9

Rank #264 Confirmed

Summary

  • They share 3 benchmarks with published results for both. Gemini 3.8 Flash scores higher in 4 categories and Qwen3-4B in 0 categories; 4 gaps are clear of the uncertainty.
  • The widest gap is in reasoning, where Gemini 3.8 Flash leads 76.9 to 19.2.
  • The biggest single-benchmark swing is Chess Puzzles: 61% for Gemini 3.8 Flash and 4% for Qwen3-4B.
  • Qwen3-4B has downloadable open weights; the other is API-only.

Side by side

Gemini 3.8 FlashQwen3-4B
ProviderGoogleAlibaba (Qwen)
Noometry Index61.831.9
Released2026-09-022025-04-29
WeightsProprietaryOpen
Context window1.05M—
Max output66K—
Input $ / M tokens$0.75—
Output $ / M tokens$3.75—
Results tracked506

Sponsored placements are available on pages like this one. Advertise on Noometry

Category by category

Coding Not comparable

Gemini 3.8 Flash: 59.2 (#15), Qwen3-4B: —

BenchmarkGemini 3.8 FlashQwen3-4B
DeepSWE73.8%—
FrontierCode41.2%—
CursorBench39.6%—
LMArena WebDev1584—
FrontierSWE19.6%—
SciCode56.6%—
WeirdML84.8%—
LMArena Coding1510—
ALE-Bench1,270—

Agentic & Tool Use Gemini 3.8 Flash leads

Gemini 3.8 Flash: 41.8 (#21), Qwen3-4B: 27.6 (#100)

BenchmarkGemini 3.8 FlashQwen3-4B
APEX-Agents64.3%—
Berkeley Function Calling Leaderboard—35.7%
Remote Labor Index5.8%—
GDP.pdf23.4%—
Vending-Bench 25,094—

Reasoning Gemini 3.8 Flash leads

Gemini 3.8 Flash: 76.9 (#5), Qwen3-4B: 19.2 (#268)

BenchmarkGemini 3.8 FlashQwen3-4B
Chess Puzzles61%4%
ARC-AGI-289.2%—
NYT Connections (extended)97.4%—
ARC-AGI-198.5%—
CritPt18.3%—
LMArena Hard Prompts1508—
Mystery Game Puzzles47%—
DTBench95.7%—
LMCA52.9%—
Surface Evolver Bench76.9%—
Epoch Capabilities Index156.71—

Math Gemini 3.8 Flash leads

Gemini 3.8 Flash: 65.3 (#28), Qwen3-4B: 29.7 (#240)

BenchmarkGemini 3.8 FlashQwen3-4B
OTIS Mock AIME 2024-202598.9%52.2%
FrontierMath (Tiers 1-3)68.4%—
FrontierMath Tier 422%—
MathArena Final-Answer Competitions—38.5%
ProofBench48%—
LMArena Math1528—

Knowledge Gemini 3.8 Flash leads

Gemini 3.8 Flash: 74.8 (#2), Qwen3-4B: 33.0 (#208)

BenchmarkGemini 3.8 FlashQwen3-4B
GPQA Diamond95.4%52.3%
Humanity's Last Exam44.5%—
SimpleQA Verified69.7%—
Vectara Hallucination Rate—5.7%
LMArena Expert1524—

Multimodal Not comparable

Gemini 3.8 Flash: 40.7 (#45), Qwen3-4B: —

BenchmarkGemini 3.8 FlashQwen3-4B
LMArena Vision1314—
Blueprint-Bench 238.6%—
Furniture Assembly31.7%—

Multilingual Not comparable

Gemini 3.8 Flash: 58.0 (#5), Qwen3-4B: —

BenchmarkGemini 3.8 FlashQwen3-4B
LMArena Non-English1491—
LMArena Chinese1554—
LMArena French1498—
LMArena German1493—
LMArena Japanese1502—
LMArena Korean1459—
LMArena Russian1515—
LMArena Spanish1485—

Instruction Following Not comparable

Gemini 3.8 Flash: 78.0 (#13), Qwen3-4B: —

BenchmarkGemini 3.8 FlashQwen3-4B
LMArena Instruction Following1490—

Long Context Not comparable

Gemini 3.8 Flash: 46.3 (#24), Qwen3-4B: —

BenchmarkGemini 3.8 FlashQwen3-4B
LMArena Longer Query1508—

Writing & Preference Not comparable

Gemini 3.8 Flash: 72.2 (#15), Qwen3-4B: —

BenchmarkGemini 3.8 FlashQwen3-4B
LMArena Text1499—
LMArena Creative Writing1492—
EQ-Bench Creative Writing1748—
LMArena Multi-Turn1501—

Frequently asked questions

Is Gemini 3.8 Flash better than Qwen3-4B?

Gemini 3.8 Flash is the stronger model overall, scoring 61.8 to 31.9 on the Noometry Index.

How many benchmarks do Gemini 3.8 Flash and Qwen3-4B share?

3 benchmarks have published results for both models. Gemini 3.8 Flash has 50 scored results on Noometry and Qwen3-4B has 6.

Related comparisons

Go deeper